Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Method and device for training ranking model

A technology for sorting models and training methods, applied in the network field, can solve the problems of low efficiency of training model parameters, and achieve the effect of improving efficiency

Active Publication Date: 2016-01-06
BEIJING QIYI CENTURY SCI & TECH CO LTD
View PDF10 Cites 22 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] Embodiments of the present invention provide a sorting model training method and device to solve the problem of low efficiency of training model parameters in the prior art, so as to improve the efficiency of training model parameters

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Method and device for training ranking model
  • Method and device for training ranking model
  • Method and device for training ranking model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0052] refer to figure 1 , which shows a flow chart of the steps of Embodiment 1 of a training method for a sorting model of the present invention, which may specifically include:

[0053] Step 101. According to the sorting model, sort the search results corresponding to the user's search request, and obtain the corresponding search sorting results;

[0054] The embodiments of the present invention can be applied to various application scenarios that provide users with search services and display related information obtained from searches to users, such as search engines such as Baidu, Google, and Yahoo, or other scenarios with search functions. The present invention is specific to The search scenarios are not limited. For ease of description, a search engine is used as an example for description below, and other application scenarios may refer to each other.

[0055] In a specific application, the search engine can obtain the content corresponding to the user's search reque...

Embodiment 2

[0096] refer to figure 2 , shows a schematic structural diagram of a ranking model training system of the present invention, which may specifically include: a search request processing module 210 , a ranking module 220 , a user behavior feedback module 230 and a central controller 240 .

[0097] Wherein, the search request processing module 210 is configured to receive a search request from a user, perform word segmentation processing on the search request, and then randomly initiate the search request to a certain sorting module 220, and send the search request from the sorting module 220 to the search request. Search sorted results are returned to the user.

[0098] The sorting module 220 is configured to receive and process the search request from the search request processing module 210. For example, it can use the word segmentation query inverted index obtained by the search request processing module 210, and then sort the query results based on the sorting model to obta...

Embodiment 3

[0103] refer to image 3 , which shows a flow chart of steps in Embodiment 2 of a training method for a sorting model of the present invention, which may specifically include:

[0104] Step 301, randomly sending a search request from a user to a sorting module;

[0105] Step 302, receiving the search ranking result corresponding to the search request from the ranking module;

[0106] Step 303, returning the search ranking results to the user;

[0107] Step 304: Determine the search click-through rate corresponding to the ranking model according to the user's click feedback on the search ranking results;

[0108] Step 305 , according to the search hit rate, update the model parameters of the ranking model according to a preset parameter update algorithm.

[0109] In an application example of the present invention, assuming that the number of sorting modules is 2, the sorting modules RSB_1 and RSB_2, the sorting model contains m parameters, and the model parameters are update...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The embodiment of the invention provides a method and a device for training a ranking model. The method comprises the following steps: ranking search results corresponding to a user search request according to the ranking model to obtain a corresponding search ranking result; determining the search click rate corresponding to the ranking model according to a click feedback of a user aiming at the search ranking result; and carrying out updating on model parameters of the ranking model by a preset parameter updating algorithm according to the search click rate. According to the method and the device, the model parameters can be automatically adjusted to adapt to a click behavior of the user when behavioral preference of the changes; and compared with the condition that a data set needs to be reestablished in the prior art, the efficiency of training the model parameters can be improved.

Description

technical field [0001] The invention relates to the field of network technology, in particular to a training method and device for a sorting model. Background technique [0002] With the development of Internet applications, search processing technology has become one of the most important applications of the Internet. For example, a search engine can search for all webpages that match the characteristics of the keyword according to the keyword entered by the user, then sort the search results according to the ranking model, and finally display the sorted search results to the user in the form of hyperlinks, and the user clicks on the corresponding The link can enter the corresponding network resource website, so as to find the required information. However, the above-mentioned search results are usually in the millions, and the most relevant search results are ranked first through the ranking model. Therefore, how to train the parameters in the ranking model to optimize th...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62G06F17/30
CPCG06F16/951G06F18/214
Inventor 胡军陈英傑王天畅
Owner BEIJING QIYI CENTURY SCI & TECH CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com